Guide

Frequent Questions

Short answers to questions that come up often.

General

Why Overplane?

Running overplane converts a directory of specs into functional, verified software.

Some reasons for using overplane as the tool of choice for this specific task are:

• Verification first: Every build raises each spec to a formal intermediate representation and checks it with the Z3 SMT solver — per spec and merged across specs — before any code is generated.
• Safety: Every agent run happens in an ephemeral local container (Docker or Podman) with the project mounted read-only; generated files are reconciled back only after a successful run.
• Multi-Agent: One vendor-neutral abstraction over Claude Code, Codex, Gemini CLI, and OpenCode. Bring your own API keys and pick the agent per project or per run.
• Caching: Content-hashed container images and content-addressable agent outputs make rebuilds incremental down to an individual spec, avoiding token waste.
• Cost reporting: Token usage and dollar cost are normalized across agent backends and reported per step and per build.

What is "verified"?

Why use this over Spec Kit?

GitHub Spec Kit is also spec-driven and multi-agent, but it is a set of slash-command prompts that run inside your existing coding agent, directly on your host. Overplane is a standalone build system. In practice that means:

• Isolation: Spec Kit agents run with full access to your machine; Overplane runs every agent in a sandboxed container.
• Verification: Spec Kit's quality gates (/speckit.clarify, /speckit.analyze, /speckit.checklist) are LLM-driven consistency reviews; Overplane additionally runs formal Z3 consistency checks on every spec.
• Caching: Spec Kit re-prompts from scratch; Overplane caches agent results so unchanged specs cost nothing to rebuild.
• Cost: Spec Kit has no built-in cost tracking; Overplane reports normalized token and dollar cost for every run.

See the full capability comparison (including AWS Kiro) on the homepage.